Computer and Channel Malfunctions in Distributed Database Systems

Northern Arizona University, College of Business Administration; Flagstaff, AZ

Four different types of distributed database management systems (DDBMS) were studied: Centralized Database and Directory (CDD), Fully Replicated Directory (FRD), Centrally Replicated Directory (CRD), and Distributed Data and Directory (DDD). Simulation models for each of these configurations were developed to capture their transaction processing characteristics as described in the previous section. The models were built as process view simulations using SIMSCRIPT II.5. Basis processes defined for each model include a job generator and separate processes for each transaction type: query, add, modify, and delete. The computers and communications channels of the system are defined as resources.

The models were designed to be as general as possible. The number of computers, and number of channels can be varied. In addition, the degree of local usage of data, the distribution of transactions between query, add, modify, and delete types, the processing speed of computers, and the channel speed, are all capable of being individually set for each computer or channel. Additional system parameters include the average job length (number of characters to be processed or moved for operations involving the processing or movement of data) and the average message length (number of characters to be processed or moved for operations involving the sending or receiving of messages).

Benefits of SIMSCRIPT II.5: Many performance characteristics such as average transaction processing time, system performance as resource saturation is approached, and the impact of computer or channel downtime on a system, are appropriately treated by simulation modeling.

Customer Quote: We would suggest that simulation modeling, using realistic parameter values for the database to be implemented, be undertaken before selecting a configuration for a DDBMS system.